For decades, the narrative around India’s tech sector has been consistent: it is the world’s back office, a powerhouse of IT services and implementation. But a quiet revolution has been brewing, aiming to shift that reputation from maintaining code to creating intelligence. The latest salvo in this battle comes from Sarvam AI.
Sarvam has quietly rolled out its "Indus" AI chat app on Android and iOS, currently operating in a waitlist-based beta phase. This isn’t just another wrapper around OpenAI’s GPT-4. Founded by Vivek Raghavan and Pratyush Kumar in 2023, Sarvam is betting on "sovereign" AI infrastructure—building the full stack from the ground up specifically for the Indian context.
With backing from heavy hitters like Lightspeed Venture Partners, Khosla Ventures, and Peak XV Partners, and fresh off a $41 million Series A round, Sarvam is making a serious play. But does Indus have what it takes to compete with global giants like Google and local rivals like Krutrim?
What makes the Indus app different from ChatGPT or Gemini?
If you have ever tried to use a western AI model for complex regional Indian tasks, you know the friction points. They often struggle with the nuances of local dialects or cultural context. This is exactly the gap Sarvam is trying to fill.
According to reports, the Indus app supports 22 Indian languages, but the real killer feature here is "code-switching." In India, conversation is rarely monolingual. We speak in a fluid mix of English and regional languages—Hinglish being the most prominent example. Indus is designed to handle this seamlessly, allowing users to switch between languages mid-sentence without confusing the AI.
Furthermore, the interface is voice-first. While text is the default in the West, voice is the primary internet on-ramp for millions of Indian users. By prioritizing voice interaction, Sarvam is targeting a demographic that Silicon Valley often treats as an edge case.
Beyond chat, the app includes robust document analysis capabilities. Users can upload PDFs and images for "knowledge extraction," a feature that suggests Sarvam is looking at productivity and enterprise use cases, not just casual chit-chat. The company is also teasing the potential for AI agents that can handle task automation, hinting at a future where the app does more than just answer questions.
How powerful are the Sarvam-30B and 105B models?
The skepticism around "indigenous" AI often boils down to one question: Is it actually new technology, or just fine-tuned LLaMA? Sarvam claims the former.
At the India AI Impact Summit 2026, the company unveiled its sovereign models: Sarvam-30B and Sarvam-105B. These are the engines likely powering the new Indus app. The "B" stands for billions of parameters, and while these numbers are smaller than GPT-4’s rumored trillion-parameter scale, size isn’t everything—especially when you specialize.
Sarvam claims its vision models actually outperform Google’s Gemini 3 Pro and OpenAI’s models on specific India-centric OCR (Optical Character Recognition) benchmarks. This is a bold claim, but it aligns with their strategy. By training specifically on Indian datasets, they can theoretically achieve higher accuracy on local documents and scripts than a larger, generalized global model could.
Even Google CEO Sundar Pichai has taken note. During his visit to the summit, he publicly praised the company, stating, "The work Sarvam has done developing local AI models… I just don’t see any impediments to that, and I think it is very, very well positioned." When the CEO of the company you are trying to disrupt gives you a nod, you are probably doing something right.
Is Sarvam winning the local battle against Krutrim?
Sarvam isn’t the only Indian player in the arena. Ola’s Krutrim has been vying for the title of India’s AI champion, but its journey has been rockier. Krutrim has faced criticism regarding the "indigenous nature" of its technology and has dealt with delays in its chip manufacturing roadmap and leadership exits.
In contrast, Sarvam’s launch seems more methodical. By releasing a tangible product with verified benchmarks—and Smartprix reviews calling their work "frontier-class"—Sarvam is positioning itself as the serious engineering alternative to Krutrim’s hype-heavy approach. The Sarvam blog emphasizes that "true AI sovereignty means building the full stack of technology from the ground up," a direct challenge to competitors relying on foreign APIs.
Between the Lines
The launch of Indus represents a critical pivot in the Indian tech economy: the shift from IT services to "intelligence arbitrage." Sarvam is proving that Indian firms can export high-value AI products rather than just the labor to maintain them. While global giants like OpenAI have a massive head start, Sarvam wins by playing a different game—owning the data residency and linguistic nuance that American labs treat as a secondary market. If they can maintain this momentum, they don’t need to beat GPT-4 globally; they just need to be the best option for the 1.4 billion people in India.
![Illustration related to Sarvam AI Indus App Launch: India's Sovereign AI [Analysis]](https://bytewire.press/wp-content/uploads/bytewire-images/2026/02/sarvam-ai-indus-app-launch-sovereign-ai-caa6723819.webp)
![Diagram related to Sarvam AI Indus App Launch: India's Sovereign AI [Analysis]](https://bytewire.press/wp-content/uploads/bytewire-images/2026/02/sarvam-ai-indus-app-launch-sovereign-ai-ee5def4a2a.webp)
